Summary
So Yong Lee argued that enterprise agents can process images and short clips, but fail across large video libraries where most video remains effectively unsearchable. She presented Jockey, built on Marengo and Pegasus, as a video agent that can search by meaning, reason across libraries, and output reports, metadata, structured data, or repurposed video.
Key Topics
5 key topics from Soyoung Lee at HumanX Amsterdam 2026. Thicker branches were mentioned more often in the talk.
The Next Frontier in Enterprise AI: Video's Compounding Value
Click a topic to open it
Key Takeaways
- Index entire video libraries, not isolated clips or frames.
- Avoid relying only on transcripts, captions, tags, or manual tagging.
- Use semantic video search to find precise moments across archives.
- Emit structured outputs so video intelligence can feed pipelines and dashboards.
- Apply library-level pattern analysis to improve advertising and content workflows.
“Because your agents can see, but it can't see at scale.”
S
Soyoung Lee
Speaker
Share
Up Next
GAJ
WatchNext in agenda
Who Owns the Future of Music?
Ghazi Shami, Andrea Czapary Martin, Jennifer Strong
ai musiccopyright
EL
WatchAlso on ai agents
Pitch Competition Semi Final #1
Eva Bandelj, Larry Li
pitch competitionai agents
MBJ
WatchAlso on ai agents
Pitch Solution Showcases
Mariam Hakobyan, Blake Marvin, Juan Huguet, Arash Emami, Ran Cherniavsky
ai agentsenterprise security